Abstract
Objective To examine influences of sociocultural and economic determinants on physical therapy (PT) utilization for older adults with rheumatoid arthritis (RA).
Methods In these annual cross-sectional analyses between 2012 and 2016, we accessed Medicare enrollment data and fee-for-service claims. The cohort included Medicare beneficiaries with RA based on 3 diagnosis codes or 2 codes plus a disease-modifying antirheumatic drug medication claim. We defined race and ethnicity and dual Medicare/Medicaid coverage (proxy for income) using enrollment data. Adults with a Current Procedural Terminology code for PT evaluation were classified as utilizing PT services. Associations between race and ethnicity and dual coverage and PT utilization were estimated with logistic regression analyses. Potential interactions between race and ethnicity status and dual coverage were tested using interaction terms.
Results Of 106,470 adults with RA (75.1% female; aged 75.8 [SD 7.3] years; 83.9% identified as non-Hispanic White, 8.8% as non-Hispanic Black, 7.2% as Hispanic), 9.6-12.5% used PT in a given year. Non-Hispanic Black (adjusted odds ratio [aOR] 0.77, 95% CI 0.73-0.82) and Hispanic (aOR 0.92, 95% CI 0.87-0.98) individuals had lower odds of PT utilization than non-Hispanic White individuals. Adults with dual coverage (lower income) had lower odds of utilization than adults with Medicare only (aOR 0.44, 95% CI 0.43-0.46). There were no significant interactions between race and ethnicity status and dual coverage on utilization.
Conclusion We found sociocultural and economic disparities in PT utilization in older adults with RA. We must identify and address the underlying factors that influence these disparities in order to mitigate them.
Rheumatoid arthritis (RA) affects around 1.3 million adults in the US1 and is associated with poor health-related quality of life, lower function, and pain.2 Disease-modifying antirheumatic drugs (DMARDs) are effective treatments for RA that lessen disease activity.3-6 However, even when disease activity is well controlled using DMARDs, people may still have functional limitations.3-6 Physical therapy (PT) and exercise interventions improve functional outcomes in individuals with RA,7-9 and a greater number of PT visits are associated with better functional improvement.10 Nevertheless, overall self-reported utilization of any PT services is low (15%).11
Socioeconomic status (SES), race, and ethnicity are social determinants of health12 that affect access to PT services. SES is a social determinant that includes numerous related factors such as income, education, and occupation. Adults with RA and lower SES demonstrated lower utilization of rehabilitation and rheumatology services than those with higher SES.11,13,14 Race and ethnicity are better represented as proxy determinants of health, as race and ethnicity are social (not biological) constructs that have been used to disempower and marginalize groups.15 Although non-Hispanic Black individuals with self-reported arthritis of any type had lower odds of reporting a rehabilitation visit than non-Hispanic White individuals,16 a comparison of PT utilization by race and ethnicity has not been conducted specifically in individuals with RA. One study found that non-White individuals with RA reported a similar number of rehabilitation visits as White individuals, but this study did not specify the specific racial or ethnic groups that the non-White category, which comprised < 7% of the sample, included.10 A thorough examination of PT use and number of visits by race and ethnicity and SES in a demographically diverse sample is needed to identify potential targets to reduce known disparities in functional and pain outcomes6,13,17-22 in individuals with RA.
In order to improve functional outcomes in adults with RA, we need to understand and address sociocultural and economic influences on PT utilization. Prior studies that examined associations of sociocultural and economic influences on PT or rehabilitation utilization in individuals with RA or any type of arthritis primarily used self reported data,10,11,14,16,23 which is prone to recall bias. The main objective of our study was to examine the influences of sociocultural and economic determinants on PT utilization in Medicare-insured older adults with RA. Medicare data provide evidence of real-world patient care and represent a large national and racially and ethnically diverse older adult patient population, allowing for a sufficiently sized sample to explore disparities in multiple racial and ethnic groups. We hypothesized that non-Hispanic Black and Hispanic individuals would be less likely to use PT services than non-Hispanic White individuals and that individuals with dual Medicare/Medicaid coverage (proxy for lower income) would be less likely to use PT services than individuals with Medicare only. We also hypothesized that race and ethnicity status and SES will interact in these models. Among users of PT services, we hypothesized that non-Hispanic Black and Hispanic individuals and individuals with dual Medicare/Medicaid coverage would have fewer PT visits than non-Hispanic White individuals and individuals with Medicare only.
METHODS
Study design. This was a cross-sectional study using annual Medicare fee-for-service claims from 2012 to 2016. To illustrate contemporary trends in PT utilization and number of visits, we examined annual billing claims for PT services in a prevalent cohort of older adults (aged ≥ 65 years) with RA from 2013 to 2016.
Data sources. We obtained Medicare data through the Master Beneficiary Summary File (MBSF), the Chronic Condition Data Warehouse (CCW), and fee-for-service claims, for services rendered through Parts A, B, and D. This study was approved through a data use agreement with the Centers for Medicare & Medicaid Services (RSCH-2019-52704) and approved by the Brown University Institutional Review Board (no. 1810002242).
Cohort identification. We established a cohort of adults > 65 years with RA and full Medicare fee-for-service coverage (Parts A, B, and D). To be identified as having RA, participants either had (1) ≥ 3 RA diagnosis codes from medical claims spread at least 7 days apart within a 1-year period, or (2) ≥ 2 RA diagnosis codes from medical claims spread at least 7 days apart within a 1-year period with ≥ 1 medication claim for a DMARD.24,25 The observational period was 2013 to 2016. Individuals entered the cohort at first RA claim. One year before their cohort entry was defined as the baseline period, going back to January 2012.
Measures.
• Key factors. We obtained data on race, ethnicity, and Medicaid status from the MBSF. We used the Research Triangle Institute (RTI) approach to classify participants as non-Hispanic White, non-Hispanic Black (or African American), Hispanic, Asian or Pacific Islander, American Indian or Alaskan Native, other, or unknown.26 The RTI modified the beneficiary race code used by the Social Security Administration through an algorithm that identifies additional beneficiaries as Hispanic or Asian or Pacific Islander based on first and last names.26 We included racial and ethnic groups that represented ≥ 5% of the overall sample. We did not consolidate groups that represented < 5% of the overall sample due to heterogeneity that exists between groups. Dual Medicare/Medicaid coverage was used as a proxy for low income.27 Medicaid is a joint federal and state program that provides health coverage to over 70 million Americans, including lower-income individuals, children, pregnant women, older adults, and individuals with disabilities.28 For dual coverage, participants were identified dichotomously (yes/no) as having either full or partial dual Medicare/Medicaid coverage for ≥ 1 month during a year (yes) or no dual coverage in any month in a year (no).
• Demographic variables, comorbidities, and DMARD use. We divided age at cohort identification into 5-year categories (65-69, 70-74, 75-79, 80-84, and ≥ 84 years) and also included sex (male or female) and census region (Northeast, South, Midwest, West) from the MBSF.29 We obtained comorbidity data from the extended CCW.30 We selected comorbidities that may influence PT utilization and are common in individuals with RA or, more broadly, in Medicare beneficiaries.31,32 We also counted the number of comorbidities based on conditions flagged according to the Elixhauser Comorbidity Index (maximum of 30 unique comorbidities), though we did not apply the weights.33 Additionally, we identified any lifetime DMARD use (conventional synthetic or biologic) based on prescription drug event claims (Part D) and supplier claims (Part B; DMARD list available from the authors upon request).
• PT utilization and number of visits. We identified PT services billed in outpatient claims using Current Procedural Terminology (CPT) codes for PT evaluation (97001, 97161, 97162, or 97163) and other PT-related services (CPT code list available from the authors upon request). Identification of PT services was restricted to 12 months after case identification. The primary outcome included PT utilization, which represented the dichotomous (yes/no) use of PT services. The rate of PT utilization was determined as the proportion of beneficiaries with a visit containing one of the above PT evaluation CPT codes within 12 months of case identification. The number of PT visits (separate dates) during an episode of PT care was identified as a secondary outcome. We defined an episode of PT care as a visit with a PT evaluation code plus any subsequent visits with ≥ 1 PT-related CPT code within 12 months after the evaluation. We categorized the number of PT visits into 4 groups: low (1-2 visits), medium (3-8 visits), high (9-18 visits), and very high (> 18 visits).34,35
Data analysis. We generated descriptive statistics for demographic and comorbidity characteristics, PT utilization, and number of PT visits for each individual calendar year from 2013 to 2016. We compared PT utilization and number of visits by race and ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic) and by dual Medicare/Medicaid coverage (yes/no) separately. To assess influence of income on racial and ethnic disparities in PT utilization and number of visits, we tested for interaction terms in logistic regression models adjusting for other patient characteristics, comorbidities, and DMARD use. Associations of race and ethnicity and dual coverage with PT utilization and number of PT visits (PT evaluation vs none; number of PT visits characterized into subgroups) were calculated as adjusted odds ratios (aORs) with 95% CI. Non-Hispanic White individuals and adults without dual coverage were the reference groups.
RESULTS
We identified 110,671 older adults with RA. In this overall sample, 80.7% identified as non-Hispanic White, 8.5% identified as non-Hispanic Black, 7% identified as Hispanic, 2.1% identified as Asian or Pacific Islander, 0.7% identified as American Indian or Alaskan Native, 0.7% identified as other, and 0.3% identified as unknown. Individuals who identified as non-Hispanic White, non-Hispanic Black, or Hispanic were included in the analyses, leaving a final cohort of 106,470 older adults with RA (Table 1). In this cohort, 83.9% identified as non-Hispanic White, 8.8% identified as non-Hispanic Black, and 7.2% identified as Hispanic. Their average age was 75.8 (SD 7.3) years, 75.1% were female, and 22.6% had dual Medicare/Medicaid coverage over the study period. The largest proportion of adults (44.7%) were from the South. DMARDs were used by 42.7% of adults. Adults had a mean Elixhauser Comorbidity Index count of 9.8, and 7.8% of adults had a mobility impairment.
Baseline characteristics of prevalent rheumatoid arthritis in Medicare enrollees by race and ethnicity and dual coverage.
Statistically significant differences (P < 0.001) were observed in the majority of baseline characteristics across racial and ethnic, and dual coverage subgroups (Table 1). A few particularly large differences were observed. Non-Hispanic White individuals had a lower proportion of adults with dual coverage and a lower proportion of women than other racial and ethnic subgroups. Non-Hispanic Black individuals had a higher proportion of adults with mobility impairments than other racial and ethnic subgroups. Further, adults with dual coverage had a higher proportion of adults with mobility impairments and a greater proportion of women than adults without dual coverage.
Within each year of case ascertainment, 9.6-12.5% of the sample used PT. Of those who used PT, the median number of visits was 8 to 9. Approximately 9% to 11% of Hispanic individuals and 10% to 13% of non-Hispanic White individuals used PT services compared to 6-10% of non-Hispanic Black individuals (Figure 1). Further, 5% to 8% of adults with dual coverage used PT services compared to 10% to 14% of adults without dual coverage (Figure 2). The median number of PT visits ranged from 7 to 10 visits across all racial, ethnic, and dual coverage subgroups (data available from the authors upon request).
Physical therapy utilization rate (%) of older adults with rheumatoid arthritis stratified by cohort entry year and race and ethnicity.
Physical therapy utilization rate (%) of older adults with rheumatoid arthritis stratified by cohort entry year and dual coverage (dual coverage representing low income).
Among those who used PT, 1064 (9.5%) had a low (1-2) number of visits, 3257 (29.1%) had 3 to 8 visits, 3922 (35%) had 9 to 18 visits, and 2967 (26.5%) had > 18 visits. The distribution of visit counts differed significantly between dual coverage groups but did not differ across racial and ethnic groups (Table 2). The most pronounced differences between dual coverage and no dual coverage groups, respectively, were in low number of visits (13.8% vs 8.7%) and high number of visits (31.1% vs 35.7%).
Number of physical therapy visits per user of physical therapy services in older adults with rheumatoid arthritis by race and ethnicity and dual coverage.
After adjustment for covariates, we observed that non-Hispanic Black (aOR 0.77, 95% CI 0.73-0.82) and Hispanic (aOR 0.92, 95% CI 0.87-0.98) individuals had lower odds of PT utilization compared to non-Hispanic White individuals (Table 3). Adults with dual coverage had lower odds of PT utilization than adults with only Medicare coverage (aOR 0.44, 95% CI 0.43-0.46). DMARD use and female sex were associated with greater odds of PT utilization, whereas mobility impairment and older age were associated with lower odds of PT utilization. PT utilization varied across geographic areas and comorbidities. Compared to living in the Midwest, living in any other region (Northeast, South, West) was associated with greater odds of PT utilization. There were no significant interactions between race and ethnicity status and dual coverage on PT utilization or number of PT visits; thus, the interaction terms were not retained in either model.
Associations between exposure variables and physical therapy utilization (any vs none) in older adults with rheumatoid arthritis.
Among adults with RA who used PT services, there were no differences in the distributions of visits across racial and ethnic subgroups. For example, non-Hispanic Black and Hispanic individuals were as likely as non-Hispanic White individuals to use a medium number of visits compared to a low number of visits. These patterns remained for high and very high number of visits compared to low number of visits (Table 4). However, adults with dual coverage had lower odds of using more PT services compared to a low number of visits (medium [aOR 0.58, 95% CI 0.48-0.71], high [aOR 0.53, 95% CI 0.43-0.64], and very high [aOR 0.56, 95% CI 0.46-0.69]). For example, compared to adults without dual coverage, adults with dual coverage had 0.58-times the odds of using a medium number of visits than a low number of visits. Adults who used DMARDs had lower odds of using a very high number of visits compared to adults who did not use DMARDs. There were a few significant associations for other demographic and clinical conditions, such as individuals living in the Northeast having greater odds of using a higher number of visits than individuals living in the Midwest, but no striking patterns.
Associations between exposure variables and number of physical therapy visits in older adults with rheumatoid arthritis.
DISCUSSION
Our study observed sociocultural and economic disparities in PT utilization for insured older adults with RA. Consistent with our hypotheses, non-Hispanic Black and Hispanic individuals were less likely to use PT services than non-Hispanic White individuals. Adults with dual Medicare/Medicaid coverage (ie, lower income) were also less likely to use PT services than adults with Medicare-only coverage. Among adults who used PT services, those with dual coverage were less likely than adults without dual coverage to use a higher number of visits compared to a low number of visits. Contrary to our hypotheses, there were no racial or ethnic disparities in number of PT visits.
Economic determinants of differential PT use were particularly large in our study. The association between dual Medicare/Medicaid coverage (proxy for economic status) and PT utilization was stronger (larger effect size) than associations of racial and ethnic status with utilization. Similarly, Cifaldi et al reported that older adults with RA and dual Medicare/Medicaid coverage had fewer outpatient visits of any type, fewer DMARD prescriptions, and worse functional outcomes than older adults with RA and Medicare alone.14 Medicaid as an additional insurance should theoretically provide additional financial support to minimize financial barriers to accessing care. Yet disparities still exist,14 suggesting that factors related to receiving Medicaid, beyond income, influence PT utilization; further, clinics often do not accept Medicaid.36 Other social determinants of health, such as education, distance from clinic, and urban vs nonurban residence or clinical location, influence utilization of other healthcare services in individuals with RA37,38 and PT utilization in adults with other health conditions.39 These determinants may also affect access to PT services in individuals with RA. Further, healthcare providers may not provide referrals40 and individuals may also not be aware of or appropriately educated on the availability or possible benefits of PT services. Future studies should examine these potential determinants, such as referral patterns, on PT use in order to address disparities in PT utilization.
Structural racism should be discussed alongside any examination of racial and ethnic disparities, as race and ethnicity are social constructs. According to Williams et al, structural racism is a system persistent throughout institutions that aids racial groups that are considered superior and oppresses racial groups that are considered inferior.15 These institutions include the healthcare system, where experiences of discrimination from healthcare providers or the healthcare system could lead to medical mistrust and subsequent decreased use or adherence in services.41-43 Another common manifestation of structural racism is racial residential segregation, which is the separation of Black and White individuals into different neighborhoods.15 Residential segregation has strongly contributed to a dearth of socioeconomic and health resources for Black individuals.15 This likely includes a lack of PT providers, resulting in PT “deserts.”44 Huber et al conducted focus groups in one such area that was located in a medically underserved, urban, and predominantly Black community.44 The participants identified several barriers to PT access, including long distance to PT clinics, lack of insurance, limited income and lost wages, and limited knowledge of PT.44 Structural racism has not been examined in the context of RA, and we do not have appropriate data in our study to directly examine structural racism. Race and ethnicity status and dual Medicare/Medicaid coverage did not interact in our study, suggesting that sociocultural factors beyond income (eg, structural racism) influenced the disparities. Future studies should seek to examine and address structural racism related to these disparities for individuals with RA.
PT use ranged from 9.6% to 12.5% in a given year in our cohort, which is similar to prior self-reported use where 15% of adults with RA used PT services in the 6 months prior to follow-up11 and 8% of adults with self reported arthritis of any type used any rehabilitation services in a given year.16 Similar self-reported use of rehabilitation services (14% in the 6 months prior to study enrollment and 41% at any time) was found in Black individuals with RA.23 Our figures may underestimate PT use for adults with RA, as patients may have used PT services for reasons unrelated to their RA. For example, a patient with RA may have used PT for hip pain after a fall that was unrelated to their RA diagnosis.
The median number of PT visits in our study was consistent from 2013 to 2016 (8-9 visits); however, there was high variability in the number of visits, reflecting the findings of a previous study.11 About 25% of our sample had > 18 visits, representing a potential overutilization of services.34 About 10% had only 1 to 2 visits, which is likely too low for clinical improvement, as the first visit consists primarily of an evaluation.
Strengths of our study include the use of claims data from a large national and racially and ethnically diverse sample of insured older adults with RA. The racial and ethnic demographics of our study cohort are representative of older adults in the US.45 Our study is the first that we know of to compare PT utilization by race and ethnicity specifically in individuals with RA.
There are some limitations to consider. Race and ethnicity were based on RTI codes, which did not allow individuals to identify as multiple races or select race and ethnicity separately. Identification inaccuracies have been observed with RTI race and ethnicity data compared to self-report, but sensitivity and specificity values for RTI were high in identifying individuals as non-Hispanic White, non-Hispanic Black, or Hispanic.46 Individuals from several racial groups, including Asian or Pacific Islander and American Indian or Alaskan Native, were excluded from the analyses owing to small sample sizes, so results cannot be generalized to these groups. We used dual Medicare/Medicaid coverage as a proxy for low income, but Medicaid can be provided for individuals for other reasons (eg, disability status) and eligibility can vary by state. However, variability in eligibility is smaller in individuals older than 65 years.27 As we used Medicare claims, we do not have access to clinical outcomes such as function. Thus, we cannot examine associations between PT utilization and clinical outcomes. We also cannot establish the disease severity of our cohort, which could affect use or number of PT visits. PT may have been prescribed for surgery (eg, joint replacement) or non-RA conditions or injuries (eg, accident). Although it was not a focus of our study, rate of DMARD use appears low (43%) in this sample. A previous study of patients with Medicare Advantage from 2005 to 2008, included in the Healthcare Effectiveness Data and Information Set, observed a rate of 63%.47 Our study examined patients with Medicare fee-for-service from a more recent observational period (2013-2016). The differences in eligibility requirements and observational periods may have resulted in the difference in DMARD use rate.
In summary, sociocultural and economic disparities in PT utilization and number of visits were observed among a national, racially and ethnically diverse, insured population of older adults with RA. Non-Hispanic Black and Hispanic individuals were less likely to use PT services than non-Hispanic White individuals. Adults with dual Medicare/Medicaid coverage (proxy for lower income) were less likely to use PT services and used a fewer number of visits than adults without dual coverage. To address these disparities in PT utilization and number of visits, we must identify and address the underlying factors that influence these disparities.
Footnotes
This project was supported by the Center on Health Services Training and Research (CoHSTAR). The funders played no role in the design, conduct, or reporting of this study.
The authors declare no conflicts of interest relevant to this article.
- Accepted for publication May 31, 2023.
- Copyright © 2023 by the Journal of Rheumatology
